{"id":{"repo_id":"cambridge","oai_identifier":"oai:www.repository.cam.ac.uk:1810/379628"},"canonical_url":"https://search.dev.ndltd.org/etd/cambridge/oai:www.repository.cam.ac.uk:1810/379628","repository":{"repo_id":"cambridge","name":"Cambridge University","base_url":"https://api.repository.cam.ac.uk/server/oai/request"},"display":{"title":"High-Throughput Ab Initio Phase Diagram Prediction","abstract":"This work treats of crystal structure prediction under a variety of different pressures and tem- peratures. It aims to do this by extending the philosophy of high-throughput ab initio random structure search (AIRSS). AIRSS is a successful technique for the prediction of crystalline and molecular structures, but in its principal form is restricted to structure prediction in the static lattice approximation. In order to extend AIRSS to predict structures at finite temperature, techniques relying on orbital-free density functional theory (DFT) and machine-learned inter- atomic potentials to calculate the forces used in phonon simulations are developed and analysed critically. Among them, machine learning is settled on as the most promising. It combines speed-ups of several orders of magnitude compared to DFT with applicability and accuracy across the periodic table. The specific flavour of machine-learned interatomic potentials used in this work is the ephemeral data-derived potential (EDDP), which relies on a training set composed of small and diverse unit cells, whose energies are learned with a physically-inspired descriptor and comparatively small neural networks. I use EDDPs to predict the phase diagrams of the heavy metals lead and polonium. In the process, insight is gained into the best strategies for developing EDDPs, and their behaviour in dynamical calculations. The effects of spin-orbit coupling on structure and melting point are analysed. I produce the first-ever rigorous prediction of the pressure-temperature phase diagram of polonium up to hundreds of GPa.","abstract_html":"This work treats of crystal structure prediction under a variety of different pressures and tem- peratures. It aims to do this by extending the philosophy of high-throughput ab initio random structure search (AIRSS). AIRSS is a successful technique for the prediction of crystalline and molecular structures, but in its principal form is restricted to structure prediction in the static lattice approximation. In order to extend AIRSS to predict structures at finite temperature, techniques relying on orbital-free density functional theory (DFT) and machine-learned inter- atomic potentials to calculate the forces used in phonon simulations are developed and analysed critically. Among them, machine learning is settled on as the most promising. It combines speed-ups of several orders of magnitude compared to DFT with applicability and accuracy across the periodic table. The specific flavour of machine-learned interatomic potentials used in this work is the ephemeral data-derived potential (EDDP), which relies on a training set composed of small and diverse unit cells, whose energies are learned with a physically-inspired descriptor and comparatively small neural networks. I use EDDPs to predict the phase diagrams of the heavy metals lead and polonium. In the process, insight is gained into the best strategies for developing EDDPs, and their behaviour in dynamical calculations. The effects of spin-orbit coupling on structure and melting point are analysed. I produce the first-ever rigorous prediction of the pressure-temperature phase diagram of polonium up to hundreds of GPa.","abstract_has_math":false,"creators":["Salzbrenner, Pascal Thomas"],"institution":"University of Cambridge","degree_name":"Doctor of Philosophy (PhD)","degree_level":"Doctoral","degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["Pickard, Christopher J"],"committee_chairs":[],"committee_members":[],"year":2024,"date_issued":"2024-06-08","date_published":"2024-06-08","updated_at":"2026-07-22T22:24:25Z","subjects":["Computational Science","Condensed Matter Physics","Machine learning","Structure prediction"],"languages":["eng"],"rights":[],"rights_urls":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/975a2bf6-d72d-419f-8eba-275dba08e161/download","https://creativecommons.org/licenses/by/4.0/"],"identifier_entries":[]},"links":{"outbound_url":"https://doi.org/10.17863/CAM.115643","outbound_label":"DOI","outbound_source":"dc:identifier.doi"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Pickard, Christopher J"]},{"key":"dc:contributor.sponsor","label":"Sponsor","values":["Trinity Hall Department of Materials Science and Metallurgy"]},{"key":"dc:creator","label":"Author","values":["Salzbrenner, Pascal Thomas"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.issued","label":"Date","values":["2024-06-08"]},{"key":"dc:publisher.institution","label":"Dc Publisher Institution","values":["University of Cambridge"]},{"key":"dc:relation.isreferencedby.uri","label":"Dc Relation Isreferencedby URI","values":["https://www.repository.cam.ac.uk/handle/1810/379628"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"dc:type.qualificationlevel","label":"Dc Type Qualificationlevel","values":["Doctoral"]},{"key":"dc:type.qualificationname","label":"Dc Type Qualificationname","values":["Doctor of Philosophy (PhD)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Computational Science","Condensed Matter Physics","Machine learning","Structure prediction"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/975a2bf6-d72d-419f-8eba-275dba08e161/download","https://creativecommons.org/licenses/by/4.0/"]},{"key":"dc:rights.embargodate","label":"Dc Rights Embargodate","values":["2026-02-13"]},{"key":"dc:rights.embargotype","label":"Dc Rights Embargotype","values":["embargo"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.doi","label":"DOI","values":["https://doi.org/10.17863/CAM.115643"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/c9cf1a4d-8bfe-4169-99b1-6d6f78d8fc12/download"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["This work treats of crystal structure prediction under a variety of different pressures and tem- peratures. It aims to do this by extending the philosophy of high-throughput ab initio random structure search (AIRSS). AIRSS is a successful technique for the prediction of crystalline and molecular structures, but in its principal form is restricted to structure prediction in the static lattice approximation. In order to extend AIRSS to predict structures at finite temperature, techniques relying on orbital-free density functional theory (DFT) and machine-learned inter- atomic potentials to calculate the forces used in phonon simulations are developed and analysed critically. Among them, machine learning is settled on as the most promising. It combines speed-ups of several orders of magnitude compared to DFT with applicability and accuracy across the periodic table. The specific flavour of machine-learned interatomic potentials used in this work is the ephemeral data-derived potential (EDDP), which relies on a training set composed of small and diverse unit cells, whose energies are learned with a physically-inspired descriptor and comparatively small neural networks. I use EDDPs to predict the phase diagrams of the heavy metals lead and polonium. In the process, insight is gained into the best strategies for developing EDDPs, and their behaviour in dynamical calculations. The effects of spin-orbit coupling on structure and melting point are analysed. I produce the first-ever rigorous prediction of the pressure-temperature phase diagram of polonium up to hundreds of GPa."]},{"key":"dc:format.checksum.md5","label":"Dc Format Checksum Md5","values":["08ad9c29ce812e1b66773bf88abae062","87eda9de84448d1f82354d60eee3eb5f"]},{"key":"dc:title","label":"Title","values":["High-Throughput Ab Initio Phase Diagram Prediction"]}]}],"canonical_facts":{"dc:contributor.advisor":["Pickard, Christopher J"],"dc:contributor.sponsor":["Trinity Hall Department of Materials Science and Metallurgy"],"dc:creator":["Salzbrenner, Pascal Thomas"],"dc:date.issued":["2024-06-08"],"dc:description.abstract":["This work treats of crystal structure prediction under a variety of different pressures and tem- peratures. It aims to do this by extending the philosophy of high-throughput ab initio random structure search (AIRSS). AIRSS is a successful technique for the prediction of crystalline and molecular structures, but in its principal form is restricted to structure prediction in the static lattice approximation. In order to extend AIRSS to predict structures at finite temperature, techniques relying on orbital-free density functional theory (DFT) and machine-learned inter- atomic potentials to calculate the forces used in phonon simulations are developed and analysed critically. Among them, machine learning is settled on as the most promising. It combines speed-ups of several orders of magnitude compared to DFT with applicability and accuracy across the periodic table. The specific flavour of machine-learned interatomic potentials used in this work is the ephemeral data-derived potential (EDDP), which relies on a training set composed of small and diverse unit cells, whose energies are learned with a physically-inspired descriptor and comparatively small neural networks. I use EDDPs to predict the phase diagrams of the heavy metals lead and polonium. In the process, insight is gained into the best strategies for developing EDDPs, and their behaviour in dynamical calculations. The effects of spin-orbit coupling on structure and melting point are analysed. I produce the first-ever rigorous prediction of the pressure-temperature phase diagram of polonium up to hundreds of GPa."],"dc:format.checksum.md5":["08ad9c29ce812e1b66773bf88abae062","87eda9de84448d1f82354d60eee3eb5f"],"dc:identifier.doi":["https://doi.org/10.17863/CAM.115643"],"dc:identifier.uri":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/c9cf1a4d-8bfe-4169-99b1-6d6f78d8fc12/download"],"dc:language":["eng"],"dc:publisher.institution":["University of Cambridge"],"dc:relation.isreferencedby.uri":["https://www.repository.cam.ac.uk/handle/1810/379628"],"dc:rights":["https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/975a2bf6-d72d-419f-8eba-275dba08e161/download","https://creativecommons.org/licenses/by/4.0/"],"dc:rights.embargodate":["2026-02-13"],"dc:rights.embargotype":["embargo"],"dc:subject":["Computational Science","Condensed Matter Physics","Machine learning","Structure prediction"],"dc:title":["High-Throughput Ab Initio Phase Diagram Prediction"],"dc:type":["Thesis"],"dc:type.qualificationlevel":["Doctoral"],"dc:type.qualificationname":["Doctor of Philosophy (PhD)"]},"updated_at":"2026-07-22T22:24:25Z"}